A web-based multilevel framework for condition monitoring of industrial equipment

(1)

A web-based multilevel framework for

condition monitoring of industrial

equipment

WJ van Blerk

orcid.org/0000-0002-3427-1525

Dissertation submitted in fulfilment of the requirements for

the degree

Master of Engineering in

Computer and

Electronic Engineering

at the North-West University

Supervisor:

Dr JC Vosloo

Graduation ceremony: May 2019

Student number: 29892139

(2)

A web-based multilevel framework for condition monitoring of industrial equipment ii

ABSTRACT

Title: A web-based multilevel framework for condition monitoring of industrial equipment

Author: WJ van Blerk Supervisor: Dr Jan Vosloo

Keywords: Condition monitoring, industrial equipment, industrial machinery, web-based framework, online framework, configurable system, cognitive load reduction

Proactive maintenance strategies aim to maintain equipment before failure and, in doing so, avoid expensive repair costs. Condition monitoring provides relevant stakeholders with information on the mechanical health status of equipment. Condition monitoring is therefore a useful tool for assisting with proactive maintenance.

Condition monitoring consists of three key steps, namely, data acquisition, data processing, and information transfer and visualisation. Existing literature focuses heavily on data processing techniques, while literature on condition monitoring information transfer and visualisation is limited. The information transfer and visualisation aspect of existing condition monitoring systems focuses mostly on the detailed data of measurements and components.

Management level personnel in industrial organisations typically use this condition monitoring information to make resource allocation decisions. These personnel are, however, responsible for numerous operations and manual investigation of each component in these organisations, which creates cognitive overload.

This study aims to improve the information transfer and visualisation to relevant stakeholders by reducing the cognitive load and increasing the data accessibility. Thereafter, the study investigates if these improvements can improve proactive maintenance in an industrial organisation by reducing the reaction times to faulty equipment.

This study develops a generic, multilevel web-based framework to visualise the processed condition monitoring data of industrial organisations. The multilevel web-based system gives a range of stakeholders remote access to the condition monitoring data to encourage timeous maintenance decisions. The system is generic and usable in numerous industries.

The system was implemented across nine operations and five system groups in a mining organisation. The system significantly reduced the cognitive load at management levels, such

(3)

A web-based multilevel framework for condition monitoring of industrial equipment iii as the organisational and operational level, while maintaining acceptable levels of cognitive load at detail levels.

The system successfully reduced the maintenance reaction times relative to the number of maintenance reactions. As with most practical implementations, there were exceptions: one of the operations showed an increase in relative reaction times. Further investigation indicated that this increase was due to a significant drop in the number of faulty equipment, causing a decreasing denominator. The system is therefore considered a success.

(4)

A web-based multilevel framework for condition monitoring of industrial equipment iv

ACKNOWLEDGEMENTS

Thanks to ETA Operations (Pty) Ltd, Enermanage, and its sister companies for the resources, time and financial assistance to complete this study.

Furthermore, I would like to thank the following people:

• My parents, for the opportunities they have given me in life and for their continued support.

• Dr JC Vosloo, Dr JN du Plessis and Dr S van Jaarsveld, for their guidance, insights and support.

• Prof. EH Mathews and Prof. M Kleingeld, for providing the opportunity to further my studies at CRCED Pretoria.

(5)

A web-based multilevel framework for condition monitoring of industrial equipment v

2.5 Summary ... 34 3 Design overview... 36 3.1 Preamble ... 36 3.2 Requirements ... 36 3.3 Development context ... 38 3.4 Functional design ... 46 3.5 Summary ... 48 4 Detailed design ... 50 4.1 Preamble ... 50 4.2 Front-end system ... 50 4.3 Database ... 60 4.4 Configuration tool ... 67 4.5 System logic ... 72 4.6 Verification ... 79 4.7 Summary ... 87 5 Results ... 89 5.1 Preamble ... 89 5.2 Implementation ... 89

5.3 Background on case study ... 89

5.4 Validation ... 90

(6)

A web-based multilevel framework for condition monitoring of industrial equipment vi

6 Conclusions and recommendations ... 111

6.1 Conclusions of the conducted work ... 111

6.2 Recommendations for future work ... 114

Reference List ... 117

(7)

A web-based multilevel framework for condition monitoring of industrial equipment vii

LIST OF FIGURES

Figure 1: Illustration of the scale of data collection in industrial organisations ... 6

Figure 2: The three stages of information processing as proposed by Atkinson and Shiffrin (Adapted from [17]) ... 8

Figure 3: The study methodology ... 10

Figure 4: Left: A handheld thermal camera; Right: An infrared thermal image of a CNC machine ... 15

Figure 5: Top: Example of time-domain vibration data; Bottom: Example of corresponding frequency-domain vibration data [34] ... 22

Figure 6: Examples of component layout diagrams in condition monitoring systems, _{... 23}

Figure 7: Example of an application of a column graph in a condition monitoring system 34₂₃ Figure 8: The basic layout of the TurbinePhD™ fleet view36_{... 24}

Figure 9: Graphic presentation used to visualise instructional efficiency [35] ... 26

Figure 10: Introducing anomalies into a harmonious pattern versus a chaotic pattern ... 30

Figure 11: Demonstration of the effect of colour on depth perception ... 31

Figure 12: Demonstration of the effect of relative element size on user attention and focus 32 Figure 13: Demonstration of the effect of misalignment on user attention and focus ... 32

Figure 14: Existing condition monitoring functionality ... 38

Figure 15: Partial database ERD of structure-related tables ... 39

Figure 16: Visual representation of an example of a mining organisational structure... 40

Figure 17: Partial database ERD of tree-structure-related tables... 40

Figure 18: Visual representation of an example of a pumping tree structure ... 41

Figure 19: Partial database ERD of user and organisational tables ... 41

Figure 20: Partial database ERD of a user and user password table ... 42

Figure 21: Partial database ERD of a user and user right tables ... 42

Figure 22: Partial database ERD of a user and applicable structure tables ... 43

Figure 23: Propagated access in a pumping tree structure of a user with access to Mine 2 43 Figure 24: Various types of data tables ... 44

Figure 25: Partial database ERD of the tag-related tables ... 44

Figure 26: Partial database ERD of data-related tables ... 45

Figure 27: DBMS node builder tool ... 46

Figure 28: Simplified systems diagram of existing infrastructure ... 46

Figure 29: Integration of the system with existing infrastructure ... 47

Figure 30: Phases of condition monitoring ... 48 Figure 31: Illustration of system adaptability to data acquisition and data processing steps 48

(8)

A web-based multilevel framework for condition monitoring of industrial equipment viii Figure 32: Illustration of the condition monitoring parameters to measure in an organisation

... 50

Figure 33: Example of the system front-end user interface ... 51

Figure 34: Typical component status blocks ... 51

Figure 35: Status block legend ... 51

Figure 36: Status block with customised tooltip text ... 52

Figure 37: Status blocks with display values ... 52

Figure 38: Organisational overview grid ... 53

Figure 39: Operational system overview grid ... 54

Figure 40: Various operational systems inherit the system statuses of their lower levels... 55

Figure 41: Tree structure node Level Up button ... 56

Figure 42: Navigation to organisational overview level using the node picker... 56

Figure 43: Navigation to operational overview level using the node picker ... 56

Figure 44: Navigation to operational system overview level using the node picker ... 57

Figure 45: Grid block with graph linked ... 57

Figure 46: Graph with a single limit ... 58

Figure 47: Graph with multiple unique limits ... 58

Figure 48: Date range and date picker ... 59

Figure 49: Date range options ... 59

Figure 50: Graph displaying data for a seven-day period ... 59

Figure 51: Graph displaying data for a 14-day period ... 60

Figure 52: Illustration of the number of duplicate nodes in an organisational structure ... 61

Figure 53: Combination of two tree structures to form operational system node links... 62

Figure 54: Combination of two tree structures to form component parameter node links ... 62

Figure 55: Condition monitoring business tree structure ... 63

Figure 56: Condition monitoring data tree structure ... 63

Figure 57: Partial database ERD of the node-link-related tables ... 64

Figure 58: Comparison between single tree structure and two-tree structure approach ... 64

Figure 59: Partial database ERD of the format tag-related tables ... 65

Figure 60: Partial database ERD of the graph and series related tables... 66

Figure 61: Basic DBMS condition monitoring configuration tool user interface ... 67

Figure 62: Left: Business tree structure node tree; Right: Data tree structure node tree ... 68

Figure 63: Create Node Link button with breadcrumbs of the selected nodes ... 68

Figure 64: Delete Node Link button with breadcrumbs of the selected nodes ... 69

Figure 65: Node link formatting configuration ... 69

Figure 66: Tag search dialog ... 69

(9)

A web-based multilevel framework for condition monitoring of industrial equipment ix

Figure 68: Node link graph configuration section ... 70

Figure 69: Node link graph only configuration ... 71

Figure 70: Graph ordering dialog... 71

Figure 71: Node link series only configuration ... 71

Figure 72: Series limit dialog ... 72

Figure 73: Extended ASP.NET MVC architecture [16]... 73

Figure 74: Process diagram of grid loading by node link selection ... 74

Figure 75: Process diagram of grid loading by node selection ... 75

Figure 76: Process diagram of grid construction ... 76

Figure 77: Process diagram of grid formatting ... 77

Figure 78: Process diagram of graph configuration for fusion charts ... 78

Figure 79: Graph presenting half-hourly average temperature values ... 80

Figure 80: Status blocks representing pump health statuses... 80

Figure 81: Parameter detail level of a deep level gold mine pump’s temperatures ... 81

Figure 82: User with full access. Left: Condition monitoring node tree, Right: User access control setup ... 82

Figure 83: User with limited access. Left: Condition monitoring node tree; Right: User access control setup ... 82

Figure 84: Left: New organisational condition monitoring business tree structure; Right: New organisational condition monitoring data tree structure ... 83

Figure 85: Top: Two nodes selected before creating a node link; Bottom: Two nodes selected after creating a node link ... 83

Figure 86: Node links on the front-end system ... 84

Figure 87: Node link with a deeper level of navigation ... 84

Figure 88: Top: Status information configuration; Bottom: Status on front-end system ... 85

Figure 89: Top: Customised tooltip text and display value configuration; Bottom: Customised tooltip text and display value on front-end system ... 85

Figure 90: Top: Graph and series configuration; Bottom: Graph and series on front-end system ... 86

Figure 91: Organisational overview level ... 92

Figure 92: Linear comparison of the number of parameters and the number of visual elements at the organisational overview level ... 92

Figure 93: Logarithmic comparison of the number of parameters and the number of visual elements at the organisational overview level ... 93

Figure 94: Pumping operational system overview level ... 93

Figure 95: Linear comparison of the average number of parameters and the average number of visual elements at operational system overview levels... 94

(10)

A web-based multilevel framework for condition monitoring of industrial equipment x Figure 96: Logarithmic comparison of the average number of parameters and the average

number of visual elements at operational system overview levels ... 94

Figure 97: Pump temperature parameter detail level ... 95

Figure 98: Linear comparison of the average number of data points and the average number of visual elements at parameter detail levels ... 96

Figure 99: Logarithmic comparison of the average number of data points and the average number of visual elements at parameter detail levels... 96

Figure 100: Maintenance reaction time measurement ... 99

Figure 101: Average reaction times for the organisation over 15 months ... 100

Figure 102: Average reaction times vs reaction count for the organisation over 15 months ... 100

Figure 103: Average reaction times relative to the reaction count for the organisation over 15 months... 101

Figure 104: Average reaction times for Operation 4 over 11 months ... 102

Figure 105: Average reaction times for Operation 9 over 15 months ... 103

Figure 106: Average reaction times vs reaction count for Operation 4 over 11 months ... 103

Figure 107: Average reaction times vs reaction count for Operation 9 over 15 months ... 104

Figure 108: Average reaction times relative to the reaction count for Operation 4 over 11 months... 104

Figure 109: Average reaction times relative to the reaction count for Operation 9 over 11 months... 105

Figure 110: Average reaction times of the various operations ... 107

(11)

A web-based multilevel framework for condition monitoring of industrial equipment xi

LIST OF TABLES

Table 1: Summary of system requirements ... 37

Table 2: Descriptions of the various status block statuses... 52

Table 3: Descriptions of the various format types ... 65

Table 4: Summary of applicable GridVM object properties ... 73

Table 5: Summary of applicable GridBlockVM object properties ... 73

Table 6: Summary of applicable GridBlockDisplayInfoVM object properties ... 74

Table 7: Summary of the amount of information that the system monitors ... 91

Table 8: Concepts applicable to the results of this subsection ... 99

Table 9: Raw data of minutes spent configuring grid blocks during Development Phase 1 123 Table 10: Raw data of minutes spent configuring format tags during Development Phase 1 ... 123

Table 11: Raw data of minutes spent configuring graphs during Development Phase 2 .... 124

Table 12: Raw data of minutes spent configuring data tags during Development Phase 3 124 Table 13: Configuration times for a partial operation during Development Phase 2 ... 125

Table 14: Configuration times for a partial operation during Development Phase 3 ... 125

(12)

A web-based multilevel framework for condition monitoring of industrial equipment xii

LIST OF ABBREVIATIONS

ARM Autoregressive modelling

BS Business structure

CNC Computer numerical control

DAL Data access layer

DB Database

DBMS Database management system

DE Drive end

DS Data structure

DSM Demand-side management

EMS Energy management system

ERD Entity relationship diagram

ESCo Energy services company

FK Foreign key

GUI Graphical user interface

IT Information technology

LDA Linear discriminant analysis

NDE Non-drive end

OPC Open platform communications

SCADA Supervisory control and data acquisition

SMS Short message service

SVM Support vector machine

(13)

A web-based multilevel framework for condition monitoring of industrial equipment xiii

NOMENCLATURE

ASP.NET MVC A version of the Microsoft®_{ASP.NET web application framework that}

uses a model-view-controller pattern.

DasBox GS&S’s condition monitoring system for industrial plant maintenance.

Fault diagnosis Identifying the root cause and severity of a mechanical fault in a piece

of equipment.

Fault prognosis Estimation of the remaining useful operational life of a faulty piece of

equipment.

FusionCharts A JavaScript library for displaying charts in web applications.

GS&S Global Solutions & Services – an Italian engineering company.

Plug and play An interface between two computer components that does not require

reconfiguration or manual installation. In terms of software, it refers to two systems that can integrate without the need for additional development.

NRG Systems A leading company in the renewable energy industry.

(14)

A web-based multilevel framework for condition monitoring of industrial equipment 1

1 Introduction

1.1 Preamble

This chapter provides background to the identified problem. Section 1.2 gives an overview of the role of maintenance in industrial organisations. Section 1.3 describes condition monitoring as a tool to assist with the maintenance of industrial equipment and the associated challenges. Section 1.4 describes the problem of cognitive overload in maintenance management and condition monitoring.

Section 1.5 formulates the problem and justifies the need for the study. Section 1.6 describes the objectives, which aim to solve the problem defined in Section 1.5. Section 1.7 describes the methodology that will be followed to achieve and validate the objectives defined in Section 1.6.

Section 1.8 provides an overview of the document.

1.2 Maintenance in the industry

1.2.1 The need for maintenance

The main driver for industrial organisations is profit. Most industrial organisations operate in either the primary (extraction of raw materials) or secondary (manufacturing and construction) economical sector. A study done on the American manufacturing industry identified that quality, cost and product lead time are the main contributing factors to a company’s competitiveness. These three factors are, however, not only applicable to the American manufacturing industry but also to the global primary and secondary sectors [1].

Production quality, cost and product lead time are all influenced by equipment reliability and maintenance. Well-maintained equipment can manufacture to finer tolerances, therefore producing higher quality products with greater consistency. This reduces the number of scrap pieces produced, which reduces costs. Well-maintained equipment increases product lead time through reduced downtime [1]. Extended downtime caused by poorly maintained equipment will not only hinder product lead time but can also have a detrimental effect on cost-saving initiatives and safety in the workplace.

Electrical cost savings projects, such as energy efficiency and demand-side management (DSM) projects, are largely dependent on the efficiency and availability of equipment [2], [3]. Multiple studies have found that optimising a maintenance strategy can lead to significant cost

(15)

A web-based multilevel framework for condition monitoring of industrial equipment 2 savings by reducing energy usage, reducing maintenance costs and increasing product lead time [4]–[6].

Equipment downtime can potentially be a major safety concern. Ventilation systems are responsible for extracting potentially harmful gases. Refrigeration systems are used to ensure safe working temperatures. Dewatering pumps on mines are responsible for flood prevention on deep levels. Downtime of any of these systems puts the health and lives of workers at risk.

Operating and maintaining industrial equipment reliably for an extended period is no simple task. Maintenance of industrial equipment accounts for 25–40% of total equipment cost (procurement, installation and operation included) depending on the type of equipment and maintenance strategy used [3], [4]. Applying overly conservative maintenance (doing maintenance too often) can lead to unnecessary maintenance costs and equipment downtime. Not applying maintenance often enough will eventually lead to equipment failure, high replacement costs and extended equipment downtime. The various types of maintenance are described in Section 1.2.2.

1.2.2 Types of maintenance

In his PhD thesis, A Performance-centered Maintenance Strategy for Industrial DSM Projects, Groenewald divides maintenance into five types, namely: breakdown maintenance, corrective maintenance, preventative maintenance, reliability-centred maintenance and total productive maintenance. Preventative maintenance is further divided into time-based and condition-based maintenance [4]. Another maintenance type, not mentioned by Groenewald, is predictive maintenance [7]. Each of these maintenance types can be categorised as either reactive or proactive maintenance. A short description of each category and maintenance type is provided in Section 1.2.2.1 and Section 1.2.2.2.

1.2.2.1 Reactive maintenance

Reactive maintenance typically occurs after equipment damage or failure. The various types of reactive maintenance are described below.

• Breakdown maintenance: Maintenance performed after equipment failure has occurred. This typically entails replacing components [4].

• Corrective maintenance: Upgrades to equipment with the goal of improving component reliability or correcting a flawed designed [4].

• Total productive maintenance: Operators are not only responsible for production, but also for reporting on maintenance needs. The aim is to maximise production while

(16)

A web-based multilevel framework for condition monitoring of industrial equipment 3 maintaining equipment reliability [4]. Operators tend to lack the skills necessary to make maintenance decisions before equipment damage occurs.

1.2.2.2 Proactive maintenance

Proactive maintenance is aimed at preventing equipment damage or failure. The various types of proactive maintenance are described below.

• Preventative maintenance: Maintenance performed with the goal of preventing equipment failure:

▪ Time-based maintenance: Maintenance done on a predetermined schedule [4]. ▪ Condition-based maintenance: Maintenance done based on the current

condition of equipment [4].

• Reliability-centred maintenance: Preventative maintenance performed only on parts that directly affect the overall system reliability [4].

• Total productive maintenance: Operators are not only responsible for production, but also for reporting on maintenance needs. The aim is maximising production while maintaining equipment reliability [4]. A properly trained and knowledgeable operator makes proactive maintenance decisions before equipment damage occurs.

• Predictive maintenance: The remaining useful life of equipment is predicted based on the current equipment health, historical data and manufacturer specifications [8]. Maintenance is done as close to the predicted date of failure as possible. The company can decide how much risk it is willing to take and how long before the predicted date of failure maintenance should be done.

Proactive maintenance tends to be more cost-effective than reactive maintenance. Proactive maintenance sacrifices equipment uptime in the short term for higher equipment reliability in the long term. The temporary loss in production rates is offset by the increase in equipment life. Proactive maintenance thus shields the company from expensive component replacement costs and extended equipment downtimes [7], [9].

When comparing the cost of general reactive maintenance with proactive maintenance types, such as preventative and predictive maintenance, the benefits of proactive maintenance become clear. A study on substation and service transformers found that reactive maintenance costs can increase the total equipment costs by up to 40%. Preventative maintenance can reduce these costs to approximately 15%. Predictive maintenance can reduce these costs further to 5–6% of the total equipment cost [7].

(17)

A web-based multilevel framework for condition monitoring of industrial equipment 4 1.2.3 Maintenance practices in the industry

The industry consists of a wide range of organisations with different backgrounds and philosophies on maintenance. Many of these organisations use reactive maintenance strategies1_{[4], [7], [10]. Section 1.2.1 noted the safety concerns of a poor maintenance}

strategy while Section 1.2.2 identified the clear benefits of proactive maintenance over reactive maintenance. Organisations that implement reactive maintenance strategies will be forced to implement proactive maintenance at some point to remain competitive.

In many cases, industrial organisations aim to implement proactive maintenance strategies such as scheduled or time-based preventative maintenance, but in practice, maintenance is done in a reactive manner [11]. Companies tend to fall back on reactive maintenance due to various factors. It is difficult to predict equipment failure without the necessary information on equipment health. This combined with the ambiguity of the financial impact of performing maintenance before failure can very easily lead to a mindset of “if it isn’t broke, don’t fix it”1_[6].

Additionally, short-term production rate goals are prioritised over temporary downtime for maintenance. In the South African mining industry, for instance, maintenance work orders are typically generated through daily inspections [10]. These inspections are done to identify equipment breakdowns or damage. Even though the mine may have a maintenance schedule in place for the equipment it is operating, it will often not adhere to the schedule as the inspections indicate that the equipment is still running.

The protocol for allocating maintenance resources, such as personnel, capital and equipment downtime, can vary from organisation to organisation; however, the basic outline is usually similar. In organisations with limited infrastructure, maintenance work orders are generated through equipment inspections, conducted on a predetermined schedule by trained personnel, or on a word-of-mouth basis where control room operators report to their supervisors if equipment behaves abnormally or breaks down [10]. Some organisations have implemented computerised maintenance management systems. These systems can be integrated with condition monitoring systems, as discussed in Section 1.3, to automate the generation of maintenance work orders and assign the applicable personnel.2

1 _{https://www.linkedin.com/pulse/key-advantages-proactive-vs-reactive-maintenance-ben-hailes/}

[Accessed February 2018]

2

(18)

A web-based multilevel framework for condition monitoring of industrial equipment 5 Maintenance work orders are escalated to the applicable organisational level depending on the severity, risk and resources required. Maintenance managers are then responsible for decision-making and resource allocation regarding maintenance.

1.3 Condition monitoring in the industry

1.3.1 The need for condition monitoring

Condition monitoring is the process of measuring and logging the operational parameters of a machine with the main purpose of observing changes that may be indicative of a potential fault before equipment failure occurs.3,4_{Vibration and temperature are the most commonly used}

parameters for determining the mechanical health of equipment through condition monitoring as they are scalar values that are relatively simple to measure [12]. Other parameters and methods that may be used for condition monitoring purposes are oil debris presence, oil pressure, motor current, ultrasonic analysis and infrared thermography5_[13].

Section 1.2.2 motivated the benefits of proactive maintenance over reactive maintenance strategies in industrial organisations. Condition monitoring is a useful tool for the implementation of preventative maintenance strategies and is crucial for the implementation of predictive maintenance strategies [4], [7], [13].

1.3.2 Challenges of industrial condition monitoring

One of the main challenges of implementing condition monitoring systems in industrial organisations is a lack of infrastructure. Large industrial organisations have been operating for decades before condition monitoring technology became available. This means that vast amounts of equipment and machinery may be operating without any form of measurement or data logging equipment installed. Implementing a condition monitoring system in such an organisation will require significant instrumentation installation resources [14].

Furthermore, these organisations may also face personnel who resist change. Uninformed, floor-level personnel may feel that their jobs are threatened by the automation of monitoring processes. Computer literacy can be very limited with floor-level and older personnel, which can lead to a reluctance to use a new system if employees feel intimidated by a complex system.

3 _{https://www.corrosionpedia.com/definition/314/condition-monitoring-cm [Accessed February 2018]}

4 _{https://www.slideshare.net/ElenaMariaVaccher/plant-maintenance-condition-monitoring [Accessed}

February 2018]

(19)

A web-based multilevel framework for condition monitoring of industrial equipment 6 Large industrial organisations can typically consist of multiple operations spread across numerous geographical regions. An operation will typically operate various systems, each consisting of multiple pieces of equipment. For effective condition monitoring, parameters should be measured at multiple points on each piece of equipment [15]. Logging all these parameters across the organisation at high resolutions can generate extremely large volumes of data. Figure 1 illustrates the scale of data collection for a condition monitoring system in an industrial organisation.

Figure 1: Illustration of the scale of data collection in industrial organisations

At control room level, simple time-domain graphs of each parameter may be sufficient to analyse and monitor equipment health. However, at operational or organisational level, this is not practical as a single organisation can measure more than 1500 parameters at a given time, as verified in Chapter 5, Section 5.4.1.2, Table 7. It is important that condition monitoring systems condense this data into useful information.

1.3.3 Requirements for industrial condition monitoring

The overarching purpose of condition monitoring systems is to provide information to the organisation and its employees. Decision makers can use this information for planning maintenance tasks, preventing potential safety risks and optimising equipment performance. It is therefore imperative that a condition monitoring system provides relevant information to the various stakeholders. For example, control room operators or vibration analysts typically need to access fine resolution data of a single piece of equipment or machinery. On the other

(20)

A web-based multilevel framework for condition monitoring of industrial equipment 7 hand, a regional maintenance manager benefits more from an overview of the operations in his/her region.6

Information is pointless if it is not accessible to those it was intended for. Due to the potential widespread nature of industrial organisations, a network-based condition monitoring system is impractical. Web-based condition monitoring systems enable users to access information regarding equipment conditions from anywhere where an internet connection is available. Additionally, the software of web-based systems tends to be simpler to maintain and does not require on-site information technology (IT) technicians for installation7_[16].

In modern times, industrial organisations often have existing IT networks, software platforms and data logging infrastructure in place. Developing and installing the necessary infrastructure for a condition monitoring system from scratch for every industrial organisation is not ideal as it is a resource-intensive process. It is therefore beneficial if a condition monitoring system can integrate seamlessly with existing infrastructure.

In terms of the scale of operations, amount of equipment/machinery and types of equipment/ machinery, there is a wide variety of industrial organisations. These organisations’ needs for a condition monitoring system are equally varying. A condition monitoring system should therefore be adaptable for various applications. The system should reflect each organisation’s needs individually in terms of information, evolution and expansion.

The rate of implementation is another key component of a condition monitoring system. If an organisation commits to using condition monitoring, the condition monitoring system should be implemented as quickly and effortlessly as possible without disrupting normal operations.

1.4 Cognitive load at management level

“Cognitive load refers to the total amount of mental effort being used in the working memory”.8

Humans have limited information processing ability and they have difficulty processing large volumes of information. A study on human memory suggests that humans process information in three stages. Figure 2 presents the three stages, which suggest that too much information will not only be hard to process and interpret, but will also be difficult to remember [17].

6

https://www.windpowerengineering.com/business-news-projects/featured/condition-monitoring-made-easy-to-read/ [Accessed September 2018]

7

http://www.differencebetween.net/technology/software-technology/difference-between-client-server-application-and-web-application/ [Accessed September 2018]

(21)

Figure 2: The three stages of information processing as proposed by Atkinson and Shiffrin (Adapted from [17])

Employees operating at management level must typically deal with complex schedules, numerous projects and resource management. The cognitive load placed on these employees is inherently high. It is thus important to avoid cognitive overload when presenting managers with additional information.

Figure 1 and Chapter 5, Section 5.4.1, give insight into the amount of information that condition monitoring system can generate. Managers need to be able to interpret this information to make well-informed maintenance and resource allocation decisions. Information condensing and simplification are essential to avoid cognitive overload for these decision makers.

1.5 Problem statement

Industrial organisations can operate vast amounts of equipment that are critical to production rates and safety in the workplace. This equipment is typically spread across numerous operations and geographical regions. Maintenance is required for continual use of this equipment and cost-effective operation of the organisation. Among costs such as procurement, installation and operating costs, maintenance can add up to 40% of the overall equipment cost [4], [7]. This leaves significant room for the optimisation of maintenance strategies to minimise costs.

Many industrial organisations tend to use reactive maintenance strategies due to various reasons [4], [7], [10]. The constant drive for high production rates, the ambiguity of the financial impact of maintenance decisions, and the difficulty of predicting remaining equipment life before breakdown are some of the main reasons for the reactive trend.

Studies have proven that proactive maintenance strategies, such as preventative and predictive maintenance, have significant long-term financial benefits over reactive maintenance [4], [7]. Condition monitoring is a useful tool to assist with preventative and

(22)

A web-based multilevel framework for condition monitoring of industrial equipment 9 predictive maintenance strategies. Through condition monitoring, the real-time mechanical condition and equipment health can be determined remotely. Maintenance managers can use this information strategically to apply maintenance before equipment failure occurs [4], [7], [13].

The need for condition monitoring varies among different industrial organisations. The size of the organisation, the organisation’s equipment, and the organisation’s financial state influence its needs for condition monitoring. A condition monitoring system therefore needs to be adaptable in the information it provides, its scalability, and its integration with existing infrastructure.

Various stakeholders typically use the system as part of the wide variety of needs that a condition monitoring system has to fulfil. These stakeholders have different job descriptions and operate at different levels within the organisation. A control room operator typically needs to access different information than an organisational-level maintenance manager.9_These

stakeholders typically operate from different locations. The system and data therefore need to be accessible remotely.

Condition monitoring at an organisational level can be challenging due to the vast amount of data that is generated. If the information from the condition monitoring system is difficult or impractical to interpret and analyse, users can experience cognitive overload [17]. This is a typical contributor to organisations reverting to reactive maintenance strategies. A condition monitoring system therefore needs to condense the available data into useful information for all relevant stakeholders.

1.6 Objectives

The objectives of the study are focused on solving the problems of condition monitoring of industrial equipment, identified in Section 1.5, by developing a web-based framework for condition monitoring. The objectives of the study are described below:

Literature review objectives

• Investigate existing solutions to industrial condition monitoring and potential shortcomings.

• Investigate solutions for cognitive overload and user interface design techniques that support cognitive load reduction.

9

(23)

A web-based multilevel framework for condition monitoring of industrial equipment 10 • Investigate framework development techniques and design principles that can assist

in the development of a customisable and configurable software system.

Main study objectives

• Develop a system that reduces the cognitive load on management level staff by condition monitoring data.

• Investigate if the system with remote accessibility and reduced cognitive load can reduce maintenance reaction times.

• Implement the system on a large multi-operational industrial organisation with as little configuration time as possible.

1.7 Methodology

Figure 3 presents the methodology with which the study aims to achieve the objectives defined in Section 1.6. A targeted literature review is conducted, which focuses on the literature review objectives defined in Section 1.6. From the literature, a set of system requirements is formulated. A detailed design of a system consisting of a front-end subsystem, a relational database and a configuration tool then commences. The design is verified regarding each of the defined requirements. Once verified, the system is implemented on a large multi-operational industrial organisation. The implementation is investigated regarding each of the main study objectives to validate if the objectives have been met.

(24)

1.8 Outline

This section provides an overview of the rest of the document.

Chapter 2 – Literature review

Presents a comprehensive review conducted on literature regarding existing work done and potential solutions to the problems identified and objectives stated in this chapter. Section 2.2 investigates existing solutions for industrial condition monitoring. Section 2.3 investigates cognitive load theory and user interface design principles to minimise cognitive load and to manage user focus. Section 2.4 reviews the advantages of modular software design.

Chapter 3 – Design overview

Gives an overview of the system design, which is aimed at solving the identified problem. The design is based on the conclusions drawn from Chapter 2. Section 3.2 formulates requirements for the system. Section 3.3 gives background of the development context and resources. Section 3.4 gives an overview of the basic operating principles and the system’s integration with existing infrastructure.

Chapter 4 – Detailed design

Describes the various elements of the system design, aimed at solving the identified problem in detail. Section 4.2 presents the front-end system functionality and features. Section 4.3 describes the relational database design. Section 4.4 describes the functionality and features of the configuration tool. Section 4.5 gives an overview of the logic flow of the system. Section 4.6 verifies that the system meets the requirements formulated in Section 3.2.

Chapter 5 – Results

Validates that the system solves the identified problem and that the system achieves the objectives of the study. Section 5.2 gives background on the implementation of the system. Section 5.3 provides background on the validation case study. Section 5.4 presents a case study where the system is implemented in a mining group. In this section, the system is evaluated in terms of each of the main study objectives.

Chapter 6 – Conclusions and recommendations

Chapter 6 concludes the study. Section 6.1 summarises the work done, discusses the case study results and concludes the success of the study. Section 6.2 identifies the shortcomings in the study and recommends future research and development work.

(25)

2 Literature review

2.1 Preamble

The chapter presents a literature review on concepts relevant to the problems and objectives mentioned in Chapter 1. The purpose of the literature review is to obtain knowledge of existing solutions to similar problems. These solutions can then be used to assist in the development of a solution to the problems addressed in this study.

Section 2.2 investigates existing condition monitoring solutions in industrial environments. Section 2.3 investigates cognitive load theory and user interface design principles to minimise cognitive load and to manage user focus. Section 2.4 reviews the advantages of framework development and modular software design.

Section 2.5 summarises the chapter and draws a conclusion from the literature review.

2.2 Existing solutions to industrial condition monitoring

2.2.1 Introduction to the aspects of condition monitoring

In his PhD thesis, Van Jaarsveld divided the condition monitoring process into three key steps. The first step is data acquisition and preparation, which is followed by operational condition assessment and, finally, information and exception reporting [18].

According to NRG Systems, a leading company in the renewable energy industry, the process used by all condition monitoring systems to convert a physical phenomenon into a recommendation to the applicable stakeholders is very similar. This process can be broken down into six basic steps, namely, data acquisition, data processing, fault detection, fault diagnosis, prognosis and, finally, the recommendation.10

Fault detection, fault diagnosis and prognosis are processes that analyse the historical data to determine the nature of equipment faults. These processes are therefore advanced versions of data processing. The recommendation step is part of information transfer from the condition monitoring system of applicable stakeholders. NRG System’s condition monitoring philosophy can therefore be condensed into data acquisition, data processing and information transfer. This aligns with the three steps identified by Van Jaarsveld [18]. This study will therefore focus on the work done on these three aspects of condition monitoring.

10

(26)

A web-based multilevel framework for condition monitoring of industrial equipment 13 Data acquisition consists of the process and the infrastructure needed to convert a physical phenomenon, such as temperature, vibration or a crack, into measured digital values and to store these digital measurements. Measurements are typically done with analogue sensors (producing outputs such as voltage), which are then converted into more meaningful digital measurements (such as °C). Work done on data acquisition in condition monitoring is discussed in further detail in Section 2.2.2.

Data processing consists of all data analysis processes that convert large volumes of data into simpler values, for example, health indicators for a component or an estimate of a component’s remaining life. With enough information, data processing can also entail identifying the cause of equipment malfunction and generating recommendations on fixing the equipment malfunction. Work done on data processing in condition monitoring is discussed in further detail in Section 2.2.3.

Information transfer or data visualisation is the way that processed data or information is presented to the applicable stakeholders. This can be achieved through automated SMSs,11

emails, reports or software. The setup and configuration of such automated notification and visualisation systems can also be considered part of the data visualisation aspect of condition monitoring. Work done on data visualisation in condition monitoring is discussed in further detail in Section 2.2.4.

2.2.2 Data acquisition

This subsection gives an overview of the data acquisition process in a condition monitoring system. Section 2.2.2.1 discusses the typical measurements needed for condition monitoring purposes. Section 2.2.2.2 gives an overview of infrastructure needed to acquire this data and the process with which it is achieved.

2.2.2.1 Parameters and measurements

In order to determine the mechanical condition of industrial equipment or components, knowledge of the physical behaviour of the equipment is required. Various parameters can be measured to achieve this. The applicable parameters for determining the mechanical condition of industrial equipment may vary depending on industry, infrastructure, available resources and application.

The most commonly used parameters for determining the mechanical health of industrial equipment are temperature and vibration. Most pieces of industrial equipment have at least

(27)

A web-based multilevel framework for condition monitoring of industrial equipment 14 one moving part during operation causing friction, heat build-up and vibration. These are also scalar measurements, which make analysis simpler [19].

Moving components, and especially rotating components, produce vibrations throughout a machine. Changes in the magnitude and frequency of these vibrations can be indicative of changes in component balance, component wear or a need for lubrication. Vibration analysis is thus a powerful tool for early fault detection in machinery. This statement is supported by the fact that numerous studies utilise vibration analysis for fault detection and diagnoses in various applications such as internal combustion engines, rolling element bearings and compressors [20]–[22].

Using vibration as an indicator of component mechanical health can be expensive. Vibration is typically measured at the bearings. Using an accelerometer is the most common method for measuring vibration [23]. These measurements are required at virtually all the bearings in the component to be meaningful. Furthermore, industrial equipment typically consists of a multitude of bearing elements. Sensor costs can quickly become expensive when multiple pieces of equipment or even operations are considered.

Temperature is an alternative to vibration as a component health identifier, which opens up cheaper measurements options to determine the mechanical health of a component [15], [24]. Temperature is a direct result of friction between moving parts. Changes in the friction forces between two moving parts will reflect in the temperature of the parts. Deviations in the operating temperatures of components can be an indication of a fault. Temperature is a flexible indicator of the mechanical component health in terms of the level of detail it provides as well as cost. For a detailed result set, sensors can be installed at all bearing elements. For a simpler result set at reduced costs, sensors can simply measure gearbox or oil temperatures [7], [24].

Temperature can be measured with a variety of methods. Thermistors, thermocouples and infrared thermography are some examples of such measurement sensors and techniques12

[15], [25]. The best-suited measurement technique depends on its application and the budget. Thermistors are temperature-dependent resistors that have non-linear behaviour. Thermistors are therefore better suited for small temperature ranges.13_{Thermocouples generate a small}

voltage when subjected to higher temperatures. Thermocouples are typically used at higher temperatures and for larger ranges.12_{With infrared thermography, a thermal camera is pointed}

at the component during operation. The thermal camera produces a colour-coded thermal

12

https://www.allaboutcircuits.com/technical-articles/introduction-temperature-sensors-thermistors-thermocouples-thermometer-ic/ [Accessed August 2018]

(28)

A web-based multilevel framework for condition monitoring of industrial equipment 15 map of the component. This technique reduces the number of sensors while producing a detailed result set of the entire component’s temperature.

Figure 4 illustrates a thermal camera and gives an example of a thermal map of a computer numerical control (CNC) milling machine. The drawback of using temperature as a health indicator is that it is not as immediate as vibration analysis. The monitored machine needs to operate stably for a while before the temperatures can indicate a developing fault.

Figure 4: Left: A handheld thermal camera;14_{Right: An infrared thermal image of a CNC machine}15

Vibration and temperature are not the only measurements that are useful for condition monitoring of industrial equipment. Oil debris content can also be a good indicator of grinding gears or mechanical parts [13]. However, depending on the type of machinery, practical real-time measurements may be challenging.

Motor current signal analysis is another method for detecting faults in rotating machinery that drive electric motors. Faults in machines, such as a defective bearing, cause deviations in motor torque and instantaneous angular speed [26].

2.2.2.2 Basic process and infrastructure

Section 2.2.2.1 mentioned numerous methods for measuring physical phenomena that can indicate the mechanical health of components. The sensors responsible for these measurements typically produce analogue outputs such as resistance or voltage.16_These

analogue values need to be converted to digital values using a digital converter. In some

16

(29)

A web-based multilevel framework for condition monitoring of industrial equipment 16 cases, such as accelerometers, sensor output voltages are too small for digital converters and require amplification first17_[27].

Digital sensor data is transmitted to an on-site supervisory control and data acquisition (SCADA) system. SCADA systems are common in industrial organisations as they enable automated control and data logging of industrial equipment.18_{The SCADA system serves as}

a central data storage unit for a site. Data is stored on the SCADA system in the form of log files [18], [28]. SCADA systems are typically configured to display the real-time measurement data to control room operators. Even though SCADA systems are typically programmed to mitigate equipment issues through automated control, control room operators can intervene and take manual control [18].

Large industrial organisations typically consist of numerous operations, each with their own SCADA systems. Directly installing condition monitoring software on each of these SCADA systems makes software maintenance and optimal setup of the condition monitoring system impractical. To simplify data processing and software maintenance, data storage needs to be centralised. An open platform communications (OPC) server can access the SCADA data and transmit data via a mobile network. A centralised data storage server receives the data from the OPC servers and translates the data to a desirable format [18], [28].

Once data has been transferred to the centralised storage facility, the condition monitoring system can access the data for processing. Data processing for condition monitoring purposes is discussed in further detail in Section 2.2.3

2.2.3 Data processing

This subsection gives an overview of the work done on the processing of condition monitoring data. Section 2.2.3.1 discusses fault detection from condition monitoring data and the diagnosis of these detected faults. Section 2.2.3.2 discusses the work done on predicting the remaining useful life of equipment.

2.2.3.1 Fault detection and diagnoses

The first and most important feature of a condition monitoring system is fault detection. All other functions of a condition monitoring system are based on fault detection. Without fault detection, equipment will run until failure, which defeats the purpose of a condition monitoring system. The system cannot diagnose a fault that it had not detected and the remaining life of

17 _{https://www.plantservices.com/assets/wp_downloads/MeasuringDisplacementUsingAccelerometers.pdf}

[Accessed August 2018]

(30)

A web-based multilevel framework for condition monitoring of industrial equipment 17 the piece of equipment cannot be estimated accurately if all information regarding equipment faults is not available. It is therefore imperative that a condition monitoring system can detect equipment faults with a high degree of accuracy and reliability.

A study conducted on the maintenance of computer-integrated manufacturing proposes that the mean-time-between-failure of manufacturing machinery can theoretically be extended indefinitely. To achieve this, development of more accurate and reliable machine performance monitoring, fault detection and fault diagnostics is required. A machine performance or condition monitoring system that can detect faults with 100% accuracy and reliability will give maintenance personnel ample time to investigate issues, correct faults, and extend machine life [1].

Unfortunately, development of a system that can detect faults with 100% accuracy and reliability is only a theoretical concept. In practice, one can only thrive to achieve a fault detection accuracy as high as the available resources and that current technology and research allow. Consequently, numerous studies have investigated various fault detection methods in a search of improved fault detection accuracy.

In a conference on the Advances in Condition Monitoring of Machinery in Non-stationary Operations, a paper investigated the accuracy and repeatability of various supervised classification methods for vibration analysis of rolling element bearings. The study concluded that simple processing methods, such as direct frequency and time-domain analyses, can accurately and reliably detect faults and indicate mechanical health of rolling element bearings. However, these methods are not suitable for diagnosing and determining the root causes of failures. Advanced data processing methods, such as narrowband envelope analysis, modulation intensity distribution, and integrated modulation intensity distribution, are better suited for diagnostic purposes [20].

Another paper from this conference analysed the SCADA temperature data of a belt conveyer system used for copper ore transportation. Due to the significant variance in system temperature with time (due to mining schedules), immediate analysis of temperature data is inaccurate. Instead, an unsupervised learning method was developed to consider longer term data. System temperatures typically form a pattern over a one-week period. Therefore, anomalies in this pattern is an indication of a developing fault [24].

A study on fault detection in gears used stochastic resonance as an advance condition monitoring method for gears. The accuracy of existing fault detection methods was improved by reducing the number of false alarms without reducing the number of correctly identified

(31)

A web-based multilevel framework for condition monitoring of industrial equipment 18 faults. The study reduced the complexity of raw signal data with two different strategies. The first strategy involved selective filtering of certain frequencies. The second strategy was to use a high pass filter at a strategic frequency [21].

An article published in the journal Mechanical Systems and Signal Processing investigated methods of classifying reciprocating compressors. The paper compared linear discriminant analysis, neural networks, support vector machines (SVMs) and extreme learning machines. It was found that non-linear classifiers (neural networks trained by Bayesian regularisation, non-linear SVMs, and extreme learning machines) performed better than their linear counter-parts and could solve more complex problems. Large volumes of historical vibration data were available for this analysis. The accuracy of machine learning techniques would reduce greatly if limited or no historical data is available [22].

Another study developed a method for high precision classification and diagnosis of bearing faults using vibration analysis. The method combined autoregressive modelling (ARM), linear discriminant analysis (LDA) and an SVM. This complex method used ARM to classify all the features in the raw vibration signal. The LDA filtered out features that were not applicable to faults. The results from the LDA were fed into the SVM to determine the state of the bearing [29].

2.2.3.2 Prognosis

Prognosis form part of advanced condition monitoring. Prognosis is a prediction of the future health of components until failure. The first requirement of prognosis is to diagnose faults accurately. If the condition monitoring system can identify a fault accurately and diagnose the severity of the fault, the system can attempt to realistically estimate the future mechanical health.19_{The operating conditions can also have a significant impact on the equipment wear}

and health deterioration. Historical data can give insight into the operating conditions of the equipment and is therefore a useful supplement for prognosis [8].

Prognosis is a powerful tool to assist with proactive maintenance. If a condition monitoring system can accurately predict the remaining life of equipment, maintenance can be done in a timeous manner without equipment failure occurring. Prognosis of industrial machinery is not a new concept and has been successful as far back as 2004 [8]. Yan, Koç and Lee used an autoregressive-moving-average model in conjunction with a maximum-likelihood technique to extrapolate historical data and determine the ideal window for maintenance [8].

19

(32)

A web-based multilevel framework for condition monitoring of industrial equipment 19 Accuracy is, a challenging area in prognosis, especially in complex machinery such as wind turbines. Prognosis in the wind turbine industry is predominantly done using vibration analysis with the occasional temperature data analysis. The study by Nie and Wang recommends that techniques such as noise inspection and oil quality monitoring are incorporated in prognosis [30].

Noise in condition monitoring signals is one of the main contributing factors to the accuracy of prognostic methods. With enough noise in vibration signals, remaining useful life prediction can become completely infeasible, especially when using the popular conventional Bayesian method. A robust method has been developed to mitigate this issue. This method uses a constrained Kalman filtering approach to try and predict the future vibration behaviour. This method has proven to be a significant improvement over the conventional Bayesian method when a significant amount of signal noise is present [31].

2.2.4 Information transfer and data visualisation

Very little academic work focuses on information transfer and the visualisation of condition monitoring data. These aspects mostly form part of the design process during the development of condition monitoring systems. This subsection therefore gives an overview of the information transfer philosophies of existing condition monitoring systems. Section 2.2.4.1 describes the platforms on which various condition monitoring systems operate. Section 2.2.4.2 analyses the way in which condition monitoring systems present data and information to stakeholders.

2.2.4.1 Platforms

Section 1.3 motivated the benefits of providing condition monitoring information to a wide range of stakeholders. These stakeholders operate from a range of physical locations. Most condition monitoring systems therefore incorporate some form of remote access to enable stakeholders to access this condition monitoring information. This is typically achieved through either client-server applications or web applications.

Client-server applications are installed on on-site computers that directly connect to the site’s networks. Condition monitoring systems that use this approach can only be accessed through the computer on which it is installed.20_{Condition monitoring systems that integrate directly}

with SCADA or control room systems are typical client-server systems [7], [24], [32], [33].

20

(33)

A web-based multilevel framework for condition monitoring of industrial equipment 20 The advantages of client-server applications are mainly associated with performance. Data processing is streamlined since the data is available directly on the same computer. The system is also only responsible for processing a smallish data set (only one site’s data). Furthermore, these systems tend to be more user-friendly due to the flexibility of window- and form-based applications.21,22

The disadvantages of client-server applications are mainly associated with software maintenance and distribution. Software updates are used for fixing bugs, maintaining compatibility and delivering new features. Having software installed on numerous widespread networks and computers makes software updates extremely challenging. In these applications, software updates need to either be installed manually from a storage device or downloaded automatically from a remote server. Manual installation from a storage device can quickly become a logistical nightmare, while automatic downloads create other challenges such as server overload [16].

Web applications are installed on a remote server with no direct connection to any site’s network. Condition monitoring systems that use this approach can only be accessed with an active internet connection through a web browser. Typical examples of such condition monitoring systems are NRG Systems’ TurbinePhD™23, 24_{and GS&S DasBox condition}

monitoring system.25, 26

Web applications offer numerous advantages over client-server applications. The first major advantage is ease of software maintenance and distribution. A single version of the application is managed and maintained. Software updates to the server’s application are available immediately to all users of the software. This ensures that all users are running the latest and most stable version of the application [16].

Ease of access is the second major advantage that web applications have over client-server applications. Web applications can be accessed from anywhere in the world if an internet connection is available. The software is not installed on the user’s computer. This means that

21 _{https://www.cogz.com/articles/web-based-cmms-systems-compared [Accessed September 2018]}

22 _{http://www.sigmaplot.co.uk/products/SIGMA_CERF/BrowserClient.pdf [Accessed September 2018]}

23

24

https://www.windpowerengineering.com/operations-maintenance/condition-monitoring-maintenance/how-is-condition-monitoring-changing-in-the-wind-industry/ [Accessed September 2018]

September 2018]

26 _{https://www.slideshare.net/ElenaMariaVaccher/gss-services-for-industries-2016 [Accessed September}

(34)

A web-based multilevel framework for condition monitoring of industrial equipment 21 the software is available immediately to the user and the user’s computer requires little computing power. This leads to web applications often being cheaper to operate.27

The disadvantages of web applications are mainly reduced performance and flexibility. A web-based condition monitoring system does not have direct access to data. Data needs to be transferred from the site to the server before data processing can commence. Furthermore, the server carries a much greater load as it must process data for all sites and the organisation28_[18].

Web applications can only be accessed through a web browser and requires an internet connection. Some remote locations do not have an internet connection and can therefore not access web applications. Browser-based user interfaces also tend to be limited in flexibility compared with window- and form-based equivalents.29

2.2.4.2 Visualisation

Condition monitoring systems measure, log and process large volumes of data to produce large volumes of information. To use a condition monitoring system successfully, this information needs to be transferred effectively to all applicable stakeholders. Information transfer is mainly done through some form of data visualisation on the user interface.

The most common raw data in condition monitoring systems is in the time domain. Some systems that use vibration analysis have data in the frequency domain. Interpretation of both time-domain and frequency-domain data is easy when presented in the form of line graphs. Most condition monitoring systems therefore use line graphs to present raw data30, 31_[7].

Figure 5 gives an example of line graphs used to present vibration data.

27 _{https://www.vstech.com/platforms/client-server-web-based/ [Accessed September 2018]}

28 _{http://www.sigmaplot.co.uk/products/SIGMA_CERF/BrowserClient.pdf [Accessed September 2018]}

29 _{https://www.cogz.com/articles/web-based-cmms-systems-compared [Accessed September 2018]}

30

A web-based multilevel framework for condition monitoring of industrial equipment